本文所涉及到的yolov5网络为6.1版本（6.0-6.2均适用）
yolov5加入注意力机制模块的三个标准步骤（适用于本文中的任何注意力机制）
1.common.py中加入注意力机制模块
2.yolo.py中增加对应的注意力机制关键字
3.yaml文件中添加相应模块
注：所有注意力机制的添加方法都是一致的，加入注意力机制是否有效的关键在于注意力机制添加的位置，本文提供两种常用常用方法。
注：需要下列所有注意力机制已经改好的代码版本及yaml文件（到手即用），请私聊我（免费）
目录
1.CBAM注意力机制
2.SE注意力机制
3.ECA注意力注意力机制
4.CA注意力注意力机制
5.SimAM注意力机制
6.ShuffleAttention注意力机制
7.CrissCrossAttention注意力机制
1.CBAM注意力机制
class ChannelAttention(nn.Module): def __init__(self, in_planes, ratio=16): super(ChannelAttention, self).__init__() self.avg_pool = nn.AdaptiveAvgPool2d(1) self.max_pool = nn.AdaptiveMaxPool2d(1) self.f1 = nn.Conv2d(in_planes, in_planes // ratio, 1, bias=False) self.relu = nn.ReLU() self.f2 = nn.Conv2d(in_planes // ratio, in_planes, 1, bias=False) self.sigmoid = nn.Sigmoid() def forward(self, x): avg_out = self.f2(self.relu(self.f1(self.avg_pool(x)))) max_out = self.f2(self.relu(self.f1(self.max_pool(x)))) out = self.sigmoid(avg_out + max_out) return out class SpatialAttention(nn.Module): def __init__(self, kernel_size=7): super(SpatialAttention, self).__init__() assert kernel_size in (3, 7), 'kernel size must be 3 or 7' padding = 3 if kernel_size == 7 else 1 self.conv = nn.Conv2d(2, 1, kernel_size, padding=padding, bias=False) self.sigmoid = nn.Sigmoid() def forward(self, x): avg_out = torch.mean(x, dim=1, keepdim=True) max_out, _ = torch.max(x, dim=1, keepdim=True) x = torch.cat([avg_out, max_out], dim=1) x = self.conv(x) return self.sigmoid(x) class CBAM(nn.Module): # CSP Bottleneck with 3 convolutions def __init__(self, c1, c2, ratio=16, kernel_size=7): # ch_in, ch_out, number, shortcut, groups, expansion super(CBAM, self).__init__() # c_ = int(c2 * e) # hidden channels # self.cv1 = Conv(c1, c_, 1, 1) # self.cv2 = Conv(c1, c_, 1, 1) # self.cv3 = Conv(2 * c_, c2, 1) # self.m = nn.Sequential(*[Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)]) self.channel_attention = ChannelAttention(c1, ratio) self.spatial_attention = SpatialAttention(kernel_size) # self.m = nn.Sequential(*[CrossConv(c_, c_, 3, 1, g, 1.0, shortcut) for _ in range(n)]) def forward(self, x): out = self.channel_attention(x) * x # print('outchannels:{}'.format(out.shape)) out = self.spatial_attention(out) * out return out以上代码需要添加在models文件夹下的common.py文件中，具体添加位置如果找不准可以选择common.py文件的最底端（最稳妥的做法，肯定不会错），或者C3模块后面（方便查找）。
第二步，需要更改models文件夹下的yolo.py文件。可以直接ctrl+F 然后查找parse_model关键字，定位到parse_model函数，你会发现有一段这样的代码
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3TR, C3Ghost, C3x]: args.insert(2, n) # number of repeats n = 1我们仅需在第1行和第8行末尾添加CBAM即可，具体做法如下
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x, CBAM): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3TR, C3Ghost, C3x, CBAM]: args.insert(2, n) # number of repeats n = 1第三步，就是最为关键的改动yaml文件了，我们以yolov5s.yaml为例进行改进，这里仅截取关键部分，未截取部分则不做改动。
第一个版本是将CBAM放在backbone部分的最末端，这样可以使注意力机制看到整个backbone部分的特征图，将具有全局视野，类似于一个小transformer结构。
# YOLOv5 v6.0 backbonebackbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 [-1, 6, C3, [256]], [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 [-1, 9, C3, [512]], [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 [-1, 3, C3, [1024]], [-1, 1, SPPF, [1024, 5]], # 9 [-1, 3, CBAM, [1024]], # 10 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 14 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 4], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 18 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 15], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 21 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 11], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 24 (P5/32-large) [[18, 21, 24], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]第二个版本是将CBAM放在backbone部分每个C3模块的后面，这样可以使注意力机制看到局部的特征，每层进行一次注意力，可以分担学习压力。
backbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 3, CBAM, [128]], # 3 [-1, 1, Conv, [256, 3, 2]], # 4-P3/8 [-1, 6, C3, [256]], [-1, 3, CBAM, [256]], [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 [-1, 9, C3, [512]], [-1, 3, CBAM, [512]], [-1, 1, Conv, [1024, 3, 2]], # 10 -P5/32 [-1, 3, C3, [1024]], [-1, 3, CBAM, [1024]], [-1, 1, SPPF, [1024, 5]], # 13 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 9], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 17 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 21 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 18], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 24 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 14], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 27 (P5/32-large) [[21, 24, 27], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]2.SE注意力机制
同理，首先将下方代码添加在models文件夹下的common.py文件中，具体添加位置如果找不准可以选择common.py文件的最底端（最稳妥的做法，肯定不会错），或者C3模块后面（方便查找）。
class SE(nn.Module): def __init__(self, c1, c2, r=16): super(SE, self).__init__() self.avgpool = nn.AdaptiveAvgPool2d(1) self.l1 = nn.Linear(c1, c1 // r, bias=False) self.relu = nn.ReLU(inplace=True) self.l2 = nn.Linear(c1 // r, c1, bias=False) self.sig = nn.Sigmoid() def forward(self, x): b, c, _, _ = x.size() y = self.avgpool(x).view(b, c) y = self.l1(y) y = self.relu(y) y = self.l2(y) y = self.sig(y) y = y.view(b, c, 1, 1) return x * y.expand_as(x)第二步，需要更改models文件夹下的yolo.py文件。可以直接ctrl+F 然后查找parse_model关键字，定位到parse_model函数，你会发现有一段这样的代码
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3TR, C3Ghost, C3x]: args.insert(2, n) # number of repeats n = 1我们仅需在第1行和第8行末尾添加SE即可，具体做法如下
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x, SE): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3TR, C3Ghost, C3x, SE]: args.insert(2, n) # number of repeats n = 1第三步，就是最为关键的改动yaml文件了，我们以yolov5s.yaml为例进行改进，这里仅截取关键部分，未截取部分则不做改动。
第一个版本是将SE放在backbone部分的最末端，这样可以使注意力机制看到整个backbone部分的特征图，将具有全局视野，类似于一个小transformer结构。
# YOLOv5 v6.0 backbonebackbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 [-1, 6, C3, [256]], [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 [-1, 9, C3, [512]], [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 [-1, 3, C3, [1024]], [-1, 1, SPPF, [1024, 5]], # 9 [-1, 3, SE, [1024]], # 10 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 14 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 4], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 18 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 15], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 21 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 11], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 24 (P5/32-large) [[18, 21, 24], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]第二个版本是将SE放在backbone部分每个C3模块的后面，这样可以使注意力机制看到局部的特征，每层进行一次注意力，可以分担学习压力。
backbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 3, SE, [128]], # 3 [-1, 1, Conv, [256, 3, 2]], # 4-P3/8 [-1, 6, C3, [256]], [-1, 3, SE, [256]], [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 [-1, 9, C3, [512]], [-1, 3, SE, [512]], [-1, 1, Conv, [1024, 3, 2]], # 10 -P5/32 [-1, 3, C3, [1024]], [-1, 3, SE, [1024]], [-1, 1, SPPF, [1024, 5]], # 13 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 9], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 17 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 21 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 18], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 24 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 14], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 27 (P5/32-large) [[21, 24, 27], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]3.ECA注意力注意力机制
同理，首先将下方代码添加在models文件夹下的common.py文件中，具体添加位置如果找不准可以选择common.py文件的最底端（最稳妥的做法，肯定不会错），或者C3模块后面（方便查找）。
class h_sigmoid(nn.Module): def __init__(self, inplace=True): super(h_sigmoid, self).__init__() self.relu = nn.ReLU6(inplace=inplace) def forward(self, x): return self.relu(x + 3) / 6class h_swish(nn.Module): def __init__(self, inplace=True): super(h_swish, self).__init__() self.sigmoid = h_sigmoid(inplace=inplace) def forward(self, x): return x * self.sigmoid(x) class CA(nn.Module): def __init__(self, inp, oup, reduction=32): super(CA, self).__init__() self.pool_h = nn.AdaptiveAvgPool2d((None, 1)) self.pool_w = nn.AdaptiveAvgPool2d((1, None)) mip = max(8, inp // reduction) self.conv1 = nn.Conv2d(inp, mip, kernel_size=1, stride=1, padding=0) self.bn1 = nn.BatchNorm2d(mip) self.act = h_swish() self.conv_h = nn.Conv2d(mip, oup, kernel_size=1, stride=1, padding=0) self.conv_w = nn.Conv2d(mip, oup, kernel_size=1, stride=1, padding=0) def forward(self, x): identity = x n, c, h, w = x.size() x_h = self.pool_h(x) x_w = self.pool_w(x).permute(0, 1, 3, 2) y = torch.cat([x_h, x_w], dim=2) y = self.conv1(y) y = self.bn1(y) y = self.act(y) x_h, x_w = torch.split(y, [h, w], dim=2) x_w = x_w.permute(0, 1, 3, 2) a_h = self.conv_h(x_h).sigmoid() a_w = self.conv_w(x_w).sigmoid() out = identity * a_w * a_h return outECA注意力机制比较特殊，不需要改动models文件夹下的yolo.py文件，可直接使用。
第三步，就是最为关键的改动yaml文件了，我们以yolov5s.yaml为例进行改进，这里仅截取关键部分，未截取部分则不做改动。
第一个版本是将ECA放在backbone部分的最末端，这样可以使注意力机制看到整个backbone部分的特征图，将具有全局视野，类似于一个小transformer结构。
# YOLOv5 v6.0 backbonebackbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 [-1, 6, C3, [256]], [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 [-1, 9, C3, [512]], [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 [-1, 3, C3, [1024]], [-1, 1, SPPF, [1024, 5]], # 9 [-1, 3, SE, [1024]], # 10 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 14 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 4], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 18 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 15], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 21 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 11], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 24 (P5/32-large) [[18, 21, 24], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]第二个版本是将ECA放在backbone部分每个C3模块的后面，这样可以使注意力机制看到局部的特征，每层进行一次注意力，可以分担学习压力。
backbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 3, SE, [128]], # 3 [-1, 1, Conv, [256, 3, 2]], # 4-P3/8 [-1, 6, C3, [256]], [-1, 3, SE, [256]], [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 [-1, 9, C3, [512]], [-1, 3, SE, [512]], [-1, 1, Conv, [1024, 3, 2]], # 10 -P5/32 [-1, 3, C3, [1024]], [-1, 3, SE, [1024]], [-1, 1, SPPF, [1024, 5]], # 13 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 9], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 17 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 21 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 18], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 24 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 14], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 27 (P5/32-large) [[21, 24, 27], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]4.CA注意力注意力机制
同理，首先将下方代码添加在models文件夹下的common.py文件中，具体添加位置如果找不准可以选择common.py文件的最底端（最稳妥的做法，肯定不会错），或者C3模块后面（方便查找）。
class ECA(nn.Module): """Constructs a ECA module. Args: channel: Number of channels of the input feature map k_size: Adaptive selection of kernel size """ def __init__(self, channel, k_size=3): super(ECA, self).__init__() self.avg_pool = nn.AdaptiveAvgPool2d(1) self.conv = nn.Conv1d(1, 1, kernel_size=k_size, padding=(k_size - 1) // 2, bias=False) self.sigmoid = nn.Sigmoid() def forward(self, x): # feature descriptor on the global spatial information y = self.avg_pool(x) # Two different branches of ECA module y = self.conv(y.squeeze(-1).transpose(-1, -2)).transpose(-1, -2).unsqueeze(-1) # Multi-scale information fusion y = self.sigmoid(y) x= x*y.expand_as(x) return x * y.expand_as(x)第二步，需要更改models文件夹下的yolo.py文件。可以直接ctrl+F 然后查找parse_model关键字，定位到parse_model函数，你会发现有一段这样的代码
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3TR, C3Ghost, C3x]: args.insert(2, n) # number of repeats n = 1我们仅需在第1行和第8行末尾添加SE即可，具体做法如下
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x, SE): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3TR, C3Ghost, C3x, SE]: args.insert(2, n) # number of repeats n = 1
第一个版本是将CA放在backbone部分的最末端，这样可以使注意力机制看到整个backbone部分的特征图，将具有全局视野，类似于一个小transformer结构。
# YOLOv5 v6.0 backbonebackbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 [-1, 6, C3, [256]], [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 [-1, 9, C3, [512]], [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 [-1, 3, C3, [1024]], [-1, 1, SPPF, [1024, 5]], # 9 [-1, 3, CA, [1024]], # 10 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 14 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 4], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 18 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 15], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 21 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 11], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 24 (P5/32-large) [[18, 21, 24], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]第二个版本是将CA放在backbone部分每个C3模块的后面，这样可以使注意力机制看到局部的特征，每层进行一次注意力，可以分担学习压力。
# YOLOv5 v6.0 backbonebackbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 3, CA, [128]], # 3 [-1, 1, Conv, [256, 3, 2]], # 4-P3/8 [-1, 6, C3, [256]], [-1, 3, CA, [256]], [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 [-1, 9, C3, [512]], [-1, 3, CA, [512]], [-1, 1, Conv, [1024, 3, 2]], # 10 -P5/32 [-1, 3, C3, [1024]], [-1, 3, CA, [1024]], [-1, 1, SPPF, [1024, 5]], # 13 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 9], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 17 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 21 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 18], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 24 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 14], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 27 (P5/32-large) [[21, 24, 27], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]
5.SimAM注意力机制
同理，首先将下方代码添加在models文件夹下的common.py文件中，具体添加位置如果找不准可以选择common.py文件的最底端（最稳妥的做法，肯定不会错），或者C3模块后面（方便查找）。
class SimAM(torch.nn.Module): def __init__(self, channels = None,out_channels = None, e_lambda = 1e-4): super(SimAM, self).__init__() self.activaton = nn.Sigmoid() self.e_lambda = e_lambda def forward(self, x): b, c, h, w = x.size() n = w * h - 1 x_minus_mu_square = (x - x.mean(dim=[2,3], keepdim=True)).pow(2) y = x_minus_mu_square / (4 * (x_minus_mu_square.sum(dim=[2,3], keepdim=True) / n + self.e_lambda)) + 0.5 return x * self.activaton(y) 第二步，需要更改models文件夹下的yolo.py文件。可以直接ctrl+F 然后查找parse_model关键字，定位到parse_model函数，你会发现有一段这样的代码
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3TR, C3Ghost, C3x]: args.insert(2, n) # number of repeats n = 1我们仅需在第1行和第8行末尾添加SimAM即可，具体做法如下
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x, SimAM): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3TR, C3Ghost, C3x, SimAM]: args.insert(2, n) # number of repeats n = 1
第一个版本是将SimAM放在backbone部分的最末端，这样可以使注意力机制看到整个backbone部分的特征图，将具有全局视野，类似于一个小transformer结构。
# YOLOv5 v6.0 backbonebackbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 [-1, 6, C3, [256]], [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 [-1, 9, C3, [512]], [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 [-1, 3, C3, [1024]], [-1, 1, SPPF, [1024, 5]], # 9 [-1, 3, SimAM, [1024]], # 10 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 14 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 4], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 18 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 15], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 21 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 11], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 24 (P5/32-large) [[18, 21, 24], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]第二个版本是将SimAM放在backbone部分每个C3模块的后面，这样可以使注意力机制看到局部的特征，每层进行一次注意力，可以分担学习压力。
# YOLOv5 v6.0 backbonebackbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 3, SimAM, [128]], # 3 [-1, 1, Conv, [256, 3, 2]], # 4-P3/8 [-1, 6, C3, [256]], [-1, 3, SimAM, [256]], [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 [-1, 9, C3, [512]], [-1, 3, SimAM, [512]], [-1, 1, Conv, [1024, 3, 2]], # 10 -P5/32 [-1, 3, C3, [1024]], [-1, 3, SimAM, [1024]], [-1, 1, SPPF, [1024, 5]], # 13 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 9], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 17 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 21 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 18], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 24 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 14], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 27 (P5/32-large) [[21, 24, 27], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]6.ShuffleAttention注意力机制
同理，首先将下方代码添加在models文件夹下的common.py文件中，具体添加位置如果找不准可以选择common.py文件的最底端（最稳妥的做法，肯定不会错），或者C3模块后面（方便查找）。
class ShuffleAttention(nn.Module): def __init__(self, channel=512,reduction=16,G=8): super().__init__() self.G=G self.channel=channel self.avg_pool = nn.AdaptiveAvgPool2d(1) self.gn = nn.GroupNorm(channel // (2 * G), channel // (2 * G)) self.cweight = torch.ones(1, channel // (2 * G), 1, 1) self.cbias = torch.ones(1, channel // (2 * G), 1, 1) self.sweight = torch.ones(1, channel // (2 * G), 1, 1) self.sbias = torch.ones(1, channel // (2 * G), 1, 1) self.sigmoid=nn.Sigmoid() @staticmethod def channel_shuffle(x, groups): b, c, h, w = x.shape x = x.reshape(b, groups, -1, h, w) x = x.permute(0, 2, 1, 3, 4) # flatten x = x.reshape(b, -1, h, w) return x def forward(self, x): b, c, h, w = x.size() #group into subfeatures x=x.view(b*self.G,-1,h,w) #bs*G,c//G,h,w #channel_split x_0,x_1=x.chunk(2,dim=1) #bs*G,c//(2*G),h,w #channel attention x_channel=self.avg_pool(x_0) #bs*G,c//(2*G),1,1 x_channel=self.cweight*x_channel+self.cbias #bs*G,c//(2*G),1,1 x_channel=x_0*self.sigmoid(x_channel) #spatial attention x_spatial=self.gn(x_1) #bs*G,c//(2*G),h,w x_spatial=self.sweight*x_spatial+self.sbias #bs*G,c//(2*G),h,w x_spatial=x_1*self.sigmoid(x_spatial) #bs*G,c//(2*G),h,w # concatenate along channel axis out=torch.cat([x_channel,x_spatial],dim=1) #bs*G,c//G,h,w out=out.contiguous().view(b,-1,h,w) # channel shuffle out = self.channel_shuffle(out, 2) return out第二步，需要更改models文件夹下的yolo.py文件。可以直接ctrl+F 然后查找parse_model关键字，定位到parse_model函数，你会发现有一段这样的代码
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3TR, C3Ghost, C3x]: args.insert(2, n) # number of repeats n = 1我们仅需在第1行和第8行末尾添加ShuffleAttention即可，具体做法如下
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x, ShuffleAttention): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3TR, C3Ghost, C3x, ShuffleAttention]: args.insert(2, n) # number of repeats n = 1
第一个版本是将ShuffleAttention放在backbone部分的最末端，这样可以使注意力机制看到整个backbone部分的特征图，将具有全局视野，类似于一个小transformer结构。
# YOLOv5 v6.0 backbonebackbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 [-1, 6, C3, [256]], [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 [-1, 9, C3, [512]], [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 [-1, 3, C3, [1024]], [-1, 1, SPPF, [1024, 5]], # 9 [-1, 3, ShuffleAttention, [1024]], # 10 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 14 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 4], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 18 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 15], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 21 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 11], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 24 (P5/32-large) [[18, 21, 24], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]第二个版本是将ShuffleAttention放在backbone部分每个C3模块的后面，这样可以使注意力机制看到局部的特征，每层进行一次注意力，可以分担学习压力。
# YOLOv5 v6.0 backbonebackbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 3, ShuffleAttention, [128]], # 3 [-1, 1, Conv, [256, 3, 2]], # 4-P3/8 [-1, 6, C3, [256]], [-1, 3, ShuffleAttention, [256]], [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 [-1, 9, C3, [512]], [-1, 3, ShuffleAttention, [512]], [-1, 1, Conv, [1024, 3, 2]], # 10 -P5/32 [-1, 3, C3, [1024]], [-1, 3, ShuffleAttention, [1024]], [-1, 1, SPPF, [1024, 5]], # 13 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 9], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 17 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 21 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 18], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 24 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 14], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 27 (P5/32-large) [[21, 24, 27], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]7.CrissCrossAttention注意力机制
同理，首先将下方代码添加在models文件夹下的common.py文件中，具体添加位置如果找不准可以选择common.py文件的最底端（最稳妥的做法，肯定不会错），或者C3模块后面（方便查找）。
def INF(B,H,W): return -torch.diag(torch.tensor(float("inf")).repeat(H),0).unsqueeze(0).repeat(B*W,1,1).cuda()class CrissCrossAttention(nn.Module): """ Criss-Cross Attention Module""" def __init__(self, in_dim, out_channels, none): super(CrissCrossAttention,self).__init__() self.query_conv = nn.Conv2d(in_channels=in_dim, out_channels=in_dim//8, kernel_size=1) self.key_conv = nn.Conv2d(in_channels=in_dim, out_channels=in_dim//8, kernel_size=1) self.value_conv = nn.Conv2d(in_channels=in_dim, out_channels=in_dim, kernel_size=1) self.softmax = nn.Softmax(dim=3) self.INF = INF self.gamma = nn.Parameter(torch.zeros(1)) def forward(self, x): m_batchsize, _, height, width = x.size() proj_query = self.query_conv(x) proj_query_H = proj_query.permute(0,3,1,2).contiguous().view(m_batchsize*width,-1,height).permute(0, 2, 1) proj_query_W = proj_query.permute(0,2,1,3).contiguous().view(m_batchsize*height,-1,width).permute(0, 2, 1) proj_key = self.key_conv(x) proj_key_H = proj_key.permute(0,3,1,2).contiguous().view(m_batchsize*width,-1,height) proj_key_W = proj_key.permute(0,2,1,3).contiguous().view(m_batchsize*height,-1,width) proj_value = self.value_conv(x) proj_value_H = proj_value.permute(0,3,1,2).contiguous().view(m_batchsize*width,-1,height) proj_value_W = proj_value.permute(0,2,1,3).contiguous().view(m_batchsize*height,-1,width) energy_H = (torch.bmm(proj_query_H, proj_key_H)+self.INF(m_batchsize, height, width)).view(m_batchsize,width,height,height).permute(0,2,1,3) energy_W = torch.bmm(proj_query_W, proj_key_W).view(m_batchsize,height,width,width) concate = self.softmax(torch.cat([energy_H, energy_W], 3)) att_H = concate[:,:,:,0:height].permute(0,2,1,3).contiguous().view(m_batchsize*width,height,height) #print(concate) #print(att_H) att_W = concate[:,:,:,height:height+width].contiguous().view(m_batchsize*height,width,width) out_H = torch.bmm(proj_value_H, att_H.permute(0, 2, 1)).view(m_batchsize,width,-1,height).permute(0,2,3,1) out_W = torch.bmm(proj_value_W, att_W.permute(0, 2, 1)).view(m_batchsize,height,-1,width).permute(0,2,1,3) #print(out_H.size(),out_W.size()) return self.gamma*(out_H + out_W) + x第二步，需要更改models文件夹下的yolo.py文件。可以直接ctrl+F 然后查找parse_model关键字，定位到parse_model函数，你会发现有一段这样的代码
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3TR, C3Ghost, C3x]: args.insert(2, n) # number of repeats n = 1我们仅需在第1行和第8行末尾添加CrissCrossAttention即可，具体做法如下
if m in (Conv, GhostConv, Bottleneck, GhostBottleneck, SPP, SPPF, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, C3new, C3new2, C3TR, C3SPP, C3Ghost, nn.ConvTranspose2d, DWConvTranspose2d, C3x, CrissCrossAttention): c1, c2 = ch[f], args[0] if c2 != no: # if not output c2 = make_divisible(c2 * gw, 8) args = [c1, c2, *args[1:]] if m in [BottleneckCSP, C3, C3new, C3new2, C3TR, C3Ghost, C3x, CrissCrossAttention]: args.insert(2, n) # number of repeats n = 1
第一个版本是将CrissCrossAttention放在backbone部分的最末端，这样可以使注意力机制看到整个backbone部分的特征图，将具有全局视野，类似于一个小transformer结构。
# YOLOv5 v6.0 backbonebackbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 1, Conv, [256, 3, 2]], # 3-P3/8 [-1, 6, C3, [256]], [-1, 1, Conv, [512, 3, 2]], # 5-P4/16 [-1, 9, C3, [512]], [-1, 1, Conv, [1024, 3, 2]], # 7-P5/32 [-1, 3, C3, [1024]], [-1, 1, SPPF, [1024, 5]], # 9 [-1, 3, CrissCrossAttention, [1024]], # 10 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 14 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 4], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 18 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 15], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 21 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 11], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 24 (P5/32-large) [[18, 21, 24], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]第二个版本是将CrissCrossAttention放在backbone部分每个C3模块的后面，这样可以使注意力机制看到局部的特征，每层进行一次注意力，可以分担学习压力。
# YOLOv5 v6.0 backbonebackbone: # [from, number, module, args] [[-1, 1, Conv, [64, 6, 2, 2]], # 0-P1/2 [-1, 1, Conv, [128, 3, 2]], # 1-P2/4 [-1, 3, C3, [128]], [-1, 3, CrissCrossAttention, [128]], # 3 [-1, 1, Conv, [256, 3, 2]], # 4-P3/8 [-1, 6, C3, [256]], [-1, 3, CrissCrossAttention, [256]], [-1, 1, Conv, [512, 3, 2]], # 7-P4/16 [-1, 9, C3, [512]], [-1, 3, CrissCrossAttention, [512]], [-1, 1, Conv, [1024, 3, 2]], # 10 -P5/32 [-1, 3, C3, [1024]], [-1, 3, CrissCrossAttention, [1024]], [-1, 1, SPPF, [1024, 5]], # 13 ]# YOLOv5 v6.0 headhead: [[-1, 1, Conv, [512, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 9], 1, Concat, [1]], # cat backbone P4 [-1, 3, C3, [512, False]], # 17 [-1, 1, Conv, [256, 1, 1]], [-1, 1, nn.Upsample, [None, 2, 'nearest']], [[-1, 6], 1, Concat, [1]], # cat backbone P3 [-1, 3, C3, [256, False]], # 21 (P3/8-small) [-1, 1, Conv, [256, 3, 2]], [[-1, 18], 1, Concat, [1]], # cat head P4 [-1, 3, C3, [512, False]], # 24 (P4/16-medium) [-1, 1, Conv, [512, 3, 2]], [[-1, 14], 1, Concat, [1]], # cat head P5 [-1, 3, C3, [1024, False]], # 27 (P5/32-large) [[21, 24, 27], 1, Detect, [nc, anchors]], # Detect(P3, P4, P5) ]